Human-in-the-loop interrupts in Agentic AI - Model Pipeline Trace

Model Pipeline - Human-in-the-loop interrupts

This pipeline shows how a machine learning model works together with a human to improve decisions. The human can interrupt the model's process to correct or guide it, making the system smarter and safer.

Data Flow - 7 Stages

1Data Input

1000 rows x 10 columns→Raw data collection from sensors and user inputs→1000 rows x 10 columns

Sensor readings and user feedback like temperature, speed, and manual labels

↓

2Preprocessing

1000 rows x 10 columns→Clean data and normalize values→1000 rows x 10 columns

Normalized temperature values between 0 and 1

↓

3Feature Engineering

1000 rows x 10 columns→Create new features from existing data→1000 rows x 15 columns

Added moving average and difference between sensor readings

↓

4Model Training

800 rows x 15 columns→Train model on training set→Trained model

Model learns to predict system status from features

↓

5Human-in-the-loop Interrupt

200 rows x 15 columns→Human reviews model predictions and interrupts if needed→Corrected predictions and feedback

Human corrects wrong predictions during testing

↓

6Model Update

Corrected predictions and feedback→Retrain or fine-tune model with human feedback→Improved model

Model adjusts weights to reduce future errors

↓

7Final Prediction

New data rows x 15 columns→Model predicts with human oversight→Predictions with confidence scores

Model outputs status with human validation

Training Trace - Epoch by Epoch

Loss
0.7 |****
0.6 |*** 
0.5 |**  
0.4 |**  
0.3 |*   
0.2 |*   
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.65	0.60	Model starts learning but with many errors
2	0.48	0.72	Loss decreases and accuracy improves
3	0.35	0.81	Model learns important patterns
4	0.28	0.86	Better predictions, fewer mistakes
5	0.22	0.90	Model converges with good accuracy

Prediction Trace - 5 Layers

Layer 1: Input Features

Layer 2: Model Prediction

Layer 3: Human-in-the-loop Interrupt

Layer 4: Model Update

Layer 5: Final Prediction

Model Quiz - 3 Questions

What is the main role of the human in the human-in-the-loop interrupt stage?

A To collect raw data

B To review and correct model predictions

C To train the model from scratch

D To preprocess the data

Key Insight

Human-in-the-loop interrupts allow the model to learn from human corrections during training and prediction. This collaboration improves accuracy and trust by combining machine speed with human judgment.